Daniel Conde, IFIC Valencia, Uncertainty estimation in Recurrent Neural Networks for Space Weather forecasting

Europe/Vienna
Seminarroom 1 and 2, PSK 3rd floor
Claudius Krause (HEPHY Vienna (ÖAW))
Description

Daniel Conde is visiting the HEPHY-ML group this week and will present his research. 

Abstract:
Uncertainty estimation in Space Weather forecasts provides critical insights for making more informed and dependable decisions to protect essential ground-based infrastructures from the adverse effects of geomagnetic disturbances. Geomagnetic storms are disturbances of the geomagnetic field caused by interactions between the solar wind and particle populations mainly in the Earth's magnetosphere. These time-varying magnetic fields induce electrical currents on long ground-based conductors that can damage power transmission grids and other critical infrastructures on Earth. This study concentrates on the uncertainty estimation in predicting the SYM-H activity index, which quantifies the strength and duration of geomagnetic storms based on ground-based geomagnetic field observations at low and mid-latitudes. Utilizing IMF data from the ACE spacecraft stationed at the L1 Lagrangian point alongside SYM-H values, we employ a long short-term memory (LSTM) neural network model to forecast the behavior and severity of geomagnetic storms several hours in advance. A significant contribution of this work is the introduction of robust methods for estimating the uncertainties associated with these predictions; this is done through the bootstrap and dropout regularization techniques already commonplace in recurrent neural networks.

Zoom connection: 

https://oeaw-ac-at.zoom.us/j/65342317590?pwd=lywPbg8pnAdq65oE9X9QWI7DqzOcUG.1